THE FUNDAMENTAL GROUP OF A p-COMPACT GROUP W. G. DWYER AND C. W. WILKERSON 1. Introduction The notion of p-compact group [10 ] is a homotopy theoretic version of the geometric or analytic notion of compact Lie group, although the homotopy theory differs from the geometry is that there are parallel theories of p-compact groups, one for each prime number p. A key feature of the theory of compact Lie groups is the relationship between centers and fundamental groups; these play off against one another, at least in the semisimple case, in that the center of the simply connected form is the fundamental group of the adjoint form. There are explicit ways to compute the center or fundamental group of a compact Lie group in terms of the normalizer of the maximal torus [1, 5.47]. For some time there has in fact been a corresponding formula for the center of a p-compact group [11 , 7.5], but in general the fundamental group has eluded analysis. The purpose of the present paper is to remedy this deficit. For any space Y , we let H Zpi(Y ) denotes lim nH i(Y ; Z=pn). Suppose that X is a connected p-compact group, with maximal torus T and torus normalizer N T [10 , x8]. It is known that the map ss1(T ) ! ss1(X) is surjective [12 , 6.11] [21 , 5.6], or equivalently that the map H Zp2(B T ) ! H Zp2(B X) is surjective. We prove the following statement. 1.1. Theorem. If X is a connected p-compact group, then the kernel of the map H Zp2BT ! H Zp2BNT is the same as the kernel of the map H Zp2BT ! H Zp2BX. Equivalently, the image of the map H Zp2BT ! H Zp2(B N T ) is (naturally) isomorphic to ss1X. 1.2. Remark. There is a proof of the corresponding statement for com- pact Lie groups which relies on the Feshbach double coset formula ____________ Date: July 24, 2006. This research was partially supported by National Science Foundation grants DMS-0204169 and DMS-0354787. The first author would like to thank the Mittag- Leffler Institute for its hospitality during the period in which this paper was completed. 1 2 W. G. DWYER AND C. W. WILKERSON (3.1). Our proof of 1.1 uses a transfer calculation (2.4) that in practice (3.4) amounts to a weak homological reflection of the double coset for- mula; we can get away with this because we have a splitting (3.5) of H Zp2(B N T ). 1.3. Remark. It is possible to derive from 1.1 a more explicit formula for ss1X; this formula is known for p odd [4, 1.7]. Let W denote the Weyl group of X. If p is odd, then ss1X is naturally isomorphic to the module of coinvariants H 0(W ; HZp2(B T )) (see 3.6). If p = 2, then up to factors which do not contribute to ss1X, the normalizer of the torus in X is derived by F2-completion from the normalizer N TG of a maximal torus TG in a connected compact Lie group G [14 , 9.13]. The image of the map H 2(B TG ; Z) ! H 2(B N TG ; Z) is isomorphic to ss1G (3.2), and so by 1.1 the tensor product of this image with Z2 is ss1X. This image can be computed from the marked reflection lattice (ss1TG , {boe, fioe}) corresponding to the root system of G [14 , 2.17, 5.5] or, after tensoring with Z2, from the marked complete reflection lattice (ss1T, {boe, fioe}) associated to X [14 , x6, x9]. The upshot is that ss1X is the quotient of ss1T = ss2 BT = H Z22BT by the Z2-submodule generated by the elements {boe}. Another way to describe this calculation is the following. For each reflection sffin the Weyl group W , let uffbe a generator over Zp of the rank 1 submodule of ss1T given by the image of (1 - sff). If p is odd let vff= uff; if p = 2, let vff= uffor uff=2, according to whether the marking of sffis trivial or non-trivial. Then ss1X is the quotient of ss1T by the Zp-span of the elements vff. See [3] for more details. Organization of the paper. In x2 we describe a transfer map in cohomol- ogy and develop some of its properties; in x3 we use these properties to prove 1.1. 2. The transfer In this section we set up the transfer machinery we need. 2.1. Definition. A fibration f : E ! B with fibre F is admissible if B is connected, H *(B; Fp) is of finite type, H *(F ; Fp) is finite dimensional, and ss1B is a finite p-group. Given an admissible fibration f : E ! B, we construct a trans- fer map f# : H *(E; Fp) ! H *(B; Fp). This is a map of modules over H*(B; Fp) which has several key properties, enumerated in the state- ments below. FUNDAMENTAL GROUP 3 2.2. Remark. It seems virtually certain that the transfer map we con- struct can be derived by duality from the homological transfer of [7, 1.1]. However, we need a property (2.4) of the transfer that is not easy to check from the point of view of [7], and so for the sake of economy of exposition we develop an ad hoc transfer that serves our immediate purposes. 2.3. Proposition. Suppose that E0 --h-! E ? ? f0?y f?y g B0 --- ! B is a homotopy fibre square in which the vertical maps are admissible fibrations. Then g*f# = f0#h*. f0 0 f A map (E0 -! B ) ! (E -! B) between fibrations over B is a map h : E0 ! E such that fh = f0. In the case B = B Z=p, the next property will sometimes allow the transfer to be calculated by "localizing to the homotopy fixed point set". f0 f 2.4. Proposition. Suppose that h is a map (E0 -! B) ! (E -! B) between admissible fibrations over B, and that S is a multiplicative subset of H *(B; Fp) such that S-1 h*: S-1 H *(E; Fp) ! S-1 H *(E0; Fp) is an isomorpism. Then the following diagram commutes: -1h* * S-1 H *(E; Fp) -S--! S-1 H (E0; Fp) ? ? S-1f#?y ?yS-1f0# S-1 H *(B; Fp) --=-! S-1 H *(B; Fp) . Next, we want to calculate the transfer in a trivial product case. If Y is a space and y 2 Y , let Yy be the component of Y containing y, and O(Y ) the Euler characteristic of Y . For our purposes, the Euler characteristic is the alternating sum of the ranks of the Fp-homology groups. 2.5. Proposition. Suppose that the projection map f : Y x B ! B is an admissible fibration. If y 2 Y , let iy : B ! Y x B be given by iy(b) = (y, b). Then for any x 2 H *(Y x B; Fp), X f# (x) = O(Yy)i*y(x) {y}2ss0Y 4 W. G. DWYER AND C. W. WILKERSON 2.6. Proposition. Suppose that f : E ! B is an admissible fibration with fibre F . Then the composite map f# f* : H *(B; Fp) ! H *(B; Fp) is multiplication by O(F ). 2.7. Stable context. In order to construct the transfer, we work in the context of S-algebras, sometimes known as structured ring spectra or A1 ring spectra, and commutative S-algebras, sometimes known as E1 ring spectra. (The symbol S stands for the sphere spectrum.) See [15 ] or [18 ] for details and examples. If k is a commutative S-algebra, we refer to algebra spectra over k as k-algebras; a reference to a k-algebra carries with it the implication that k is a commutative S-algebra. The sphere S is itself a commutative S-algebra, and, as the term "S-algebra" suggests, any ring spectrum is an algebra spectrum over S. We refer to a module spectrum over an S-algebra R as simple a module over R. We write R for the smash product over R of two R-modules, and we always take this smash product in the appropriate derived sense. Similarly, we use Hom R for the derived spectrum of maps between two R-modules. Any ring R gives rise to an S-algebra (whose homotopy is R, concen- trated in degree 0). The category of modules over this S-algebra is then equivalent in an appropriate homotopy theoretic sense to the cagtegory of chain complexes over R. If R is commutative in the usual sense then its corresponding S-algebra is also commutative; the category of al- gebras over this S-algebra is then closely related to the category of differential graded algebras over R [22 ]. For the rest of this section, k will denote the commutative S-algebra corresponding to Fp. 2.8. Cochain algebras. We will need the following observations from [19 ] and [20 ]. For any space X, the the mapping spectrum kX = Hom S( 1 X+ , k) has the natural structure of a commutative k-algebra; this k-algebra is a geometric form of the Fp-cochain algebra of X, and in particuar ssi(kX ) = H -i(X; Fp). Suppose that X --- ! E ? ? ? ? y y Y --- ! B is a homotopy fibre square in which B is connected, and let F be the homotopy fibre of E ! B. Then if B, F , and Y have Fp-cohomology rings of finite type and B is simply connected (more generally, ss1B acts nilpotently on each Fp-homology group of F ) the natural map kY kB kE ! kX FUNDAMENTAL GROUP 5 is a weak equivalence of commutative k-algebras. Suppose that R is a commutative S-algebra and that N is an R- module. 2.9. Definition. An R-module M is finitely built from N if it is in the smallest thick subcategory of R-modules which contains N and its (de)suspensions. An R-algebra A is finite if A is finitely built from R as an R-module. More explicitly, M is finitely built from N if, up to equivalence and retracts, M can be built via cofibration sequences from finitely many copies of (de)suspensions of N. Recall that k denotes the S-algebra derived from Fp. 2.10. Lemma. If E ! B is an admissible fibration with fibre F , then kE is a finite kB -algebra. Proof. If X is a space, let k[X] = k S 1 (X+ ); this is a geometric form of the Fp-chains on X, in that ssik[X] = H i(X; Fp). Let G be the loop group B; then k[G] is a k-algebra and monodromy in the fibration E ! B makes k[F ] a module over k[G]. The Rothenberg-Steenrod construction produces an equivalence k k[G]k[F ] ~ k[E]. Dualizing gives kE ~ Hom k(k[E], k) ~ Hom k(k k[G]k[F ], k) ~ Hom k[G](k[F ], k) . Now k[F ] has only a finite number of nonzero homotopy groups, each one of which is a finite dimensional Fp-vector space which up to fil- tration is trivial as a module over ss0k[G] = Fp[ss1B]. (This last fol- lows from the fact that the augmentation ideal in Fp[ss1G] is nilpo- tent.) As in [8, 3.2, 3.9], it follows that k[F ] is finitely built from the trivial module k as a module over k[G]. Applying the construction U(M) = Hom k[G](M, k), which preserves cofibration sequences in M, shows that kE ~ U(k[F ]) is finitely built from kB ~ U(k) ~ End k[G](k) as a module over kB . The reader might be worried that there are two potential module structures above for kE as a module over kB , one coming from the natural k-algebra map kB ! kE , and the other from the expressions kE = Hom k[G](k[F ], k), kB = Hom k[G](k, k). We have proved that in the second module structure, kE is finitely built from kB . However, the two module structures are the same. We indicate very briefly how to show this; for simplicity we assume that E is connected. Let P be the path fibration over B. There is a fibrewise action of G on P which by naturality gives a map kB ! End k[G](kP ) ~ End k[G](k). This is the equivalence mentioned above; note that it is evidently a map of 6 W. G. DWYER AND C. W. WILKERSON S-algebras. If L = E, there is a similar equivalence kE ! End k[L](k), as well as a commutative diagram kB --- ! End k[G](k) ? ? ? ? y y kE --- ! End k[L](k) in which the left vertical map is the natural one and the right vertical map is induced by L ! G. By a form of Shapiro's lemma, there are equivalences kE ~ End k[L](k) ~ Hom k[G](k[G=L], k) ~ Hom k[G](k[F ], k) of modules over End k[G](k) ~ kB . If R is a commutative S-algebra, and M is a R-module, write DR M for the R-module Hom R(M, R). For any R-module N, composition gives a map N R DR (M) ~ Hom R(R, N) R Hom R(M, R) ! Hom R(M, N) 2.11. Lemma. If M is finitely built from R, then for any R-module N the natural map N R DR M ! Hom R(M, N) is an equivalence. Proof. The statement is clearly true if M is a suspension of R and fol- lows in general by an induction on the number of cofibration sequences needed to construct M from R. The key point is that both of the con- structions N R DR M and Hom R(M, N) preserve cofibration sequences in M. Suppose that M is an R-module which is finitely built from R. There is a map j = jM : R ! M R DR M corresponding to the identity map in Hom R(M, M), as well as an evaluation map ffl = fflM : M R DR M ! R. Both of these are R-module maps. 2.12. Definition. Suppose that A is a finite R-algebra. The transfer associated to A is the R-module map trA=R : A ! R given by the following composite id jA mult id fflA A ~ A R R --- - !A R A R DR A --- - - !A R DR A -! R. 2.13. Remark. If R is a commutative ring and A is an R-algebra which is finitely generated and free as an R-module, then trA=R assigns to each element a 2 A the trace over R of multiplication by a on A. 2.14. Definition. Suppose that f : E ! B is an admissible fibra- tion, so that kE is a finite kB -algebra (2.10). The transfer map f# : FUNDAMENTAL GROUP 7 H *(E; Fp) ! H *(B; Fp) is defined to be the map on homotopy groups induced by trkE=kB: kE ! kB . 2.15. Proposition. Suppose that A is a finite R-algebra, R ! R0 is a map of commutative S-algebras, and A0= R0 R A. Then A0 is a finite R0-algebra, and the following diagram commutes up to homotopy: trA=R A --- ! R ? ? ? ? y y trA0=R0 A0 --- - !R0. Proof. The finiteness condition is easy to check. The rest follows from the fact that chain of maps (2.12) which determines trA0=R0is obtained from the chain of maps which determines trA=R by applying the functor - R R0. Proof of 2.3. This is a consequence0of 2.15; since0f is admissible, it follows from 2.8 that kE is equivalent to kB kB kE . Proof of 2.4. According to [16 ], for any commutative S-algebra R and subset S of ss*R, there is a localized commutative S-algebra S-1 R to- gether with a map R ! S-1 R inducing an isomorphism S-1 ss*R ~= ss*(S-1 R). Moreover, if M is an R-module and S-1 M is defined as (S-1 R) R M, then ss*(S-1 M) ~= S-1 ss*M. The hypotheses imply 0 that S-1 kE ! S-1 kE is an equivalence, and so the proposition can be proved by two applications of 2.15. (The first application, for instance, identifies S-1 f# with the transfer associated to the S-1 (kB )-algebra given by S-1 (kE ).) Proof of 2.5. We leave it to the reader to check this in the case in which B is a point; it amounts to calcuating a composite H *(Y ; Fp)! H *(Y ; Fp) Fp H*(Y ; Fp) Fp H*(Y ; Fp) ! H *(Y ; Fp) Fp H*(Y ; Fp) ! Fp in which the only nonzero component is in degree 0. Since f# is a map of modules over H *(B; Fp), the general case can be obtained by applying 2.3 to the fibre square Y x B -- - ! Y ? ? ? ? y y . B -- - ! * 8 W. G. DWYER AND C. W. WILKERSON Proof of 2.6. Since f# f* is a map of modules over H *(B; Fp), it is enough to calculate its effect in degree 0. This can be done by applying 2.5 to the fibre square F --- ! E ? ? ? ? y y . * --- ! B 3. Using the transfer We first sketch the proof of a result parallel to 1.1 for compact Lie groups. We will have to make a few adjustments to the Lie group argu- ment in order to compensate for missing components of the homotopy theoretic machinery, but this proof is the prototype for our approach. 3.1. Theorem. Suppose that G is a connected compact Lie group with maximal torus T and torus normalizer N T . Then the kernel of the map H *(B T ; Z) ! H *(B G; Z) is the same as the kernel of the map H *(B T ; Z) ! H *(B N T ; Z). 3.2. Remark. Since H 2(B T ; Z) ! H 2(B G; Z) is surjective [13 , 9.3], this theorem calculates ss1G = H 2(B G; Z) in terms of the image of the map H 2(B T ; Z) ! H 2(B N T ; Z). 3.3. Proof of 3.1 (sketch). If f : E ! B is a fibre bundle with a compact manifold as fibre, we let f! : H *(B; Z) ! H *(E; Z) denote the associated Becker-Gottlieb-Dold transfer map [5] [6] (this is very much the same kind of map as the Fp-cohomology transfer discussed in x2). Consider the following commutative diagram of vertical fibre sequences. Here the space V is defined so that the lower right square is a homotopy fibre square; equivalently, V is the Borel construction of the left action of T on G= NT . (G= NT )T --- ! G= NT --=-! G= NT ? ? ? ? ? ? y y y (G= NT )T x BT --a-! V --b-! B N T ? ? ? u?y v?y w?y B T --=-! B T --c-! B G By a basic property (cf. 2.3) of the transfer, b*v! = w!c*. The space G= NT has Euler characteristic one, and so w*w!is the identity map of H *(B G; Z) (cf. 2.6); in particular w! is injective, and so we conclude that the kernel of c* is the same as the kernel of b*v!. FUNDAMENTAL GROUP 9 For any closed subgroup K T , let p(K) be the projection map in the fibration p(K) T =K ! BK -- ! BT . Since T is abelian, the connected compact Lie group T =K acts freely on B K; in particular, if K 6= T the fibre bundle B K ! B T admits a fibrewise self-map which is fixed-point free but fibrewise homotopic to the identity. It follows from a theorem of Dold [6] that if K 6= T the transfer map p(K)!is trivial. By the Feshbach formula [17 ], then, b*v!= b*a*u!. But the action of T on G= NT has only a single fixed point; this follows from the fact that any element of G which conjugates T into N T must conjugate T into the identity component of N T , i.e., into T itself, and hence must lie in N T . It is now clear that the composite b*a*u! is just the homology homomorphism i* induced by the usual map i :B T ! B NT , and so ker(i*) = ker(b*a*u!) = ker(b*v!) and as above this last is the same as ker(c*). For the rest of this section, X by default denotes a connected p- compact group with maximal torus T , torus normalizer N T , and Weyl group W . In addition, N pT denote the p-compact group whose classi- fying space is the cover of B NT corresponding to a Sylow p-subgroup Wp of W [10 , 9.8]. 3.4. Remark. There are two technical issues which make our proof of 1.1 slightly more complicated than the sketch above. First of all, fibra- tions with fibre X= NT are awkward to handle, because N T is not a p-compact group. Instead we work with X= NpT , a change which intro- duces some additional bookkeeping. Secondly, in the p-compact case we do not have a Feshbach transfer formula for the fibration over B T with X= NpT as the fibre, much less a Dold-type result which would focus the transfer calculation on the (homotopy) fixed points of the action of T on X= NpT . To compensate, we pull the fibration back over various maps B Z=p B T and use 2.4, which in light of Smith theory [9] does in fact allow the transfer to be calculated by restricting to (X= NpT )hZ=p. This requires dealing with Fp-(co)homology instead of with H Zp*, but there is an a priori splitting theorem (3.5) which permits this. We first handle the splitting. 3.5. Lemma. The image of the map H Zp2BT ! H Zp2BN T is a split summand of H Zp2BNT . 10 W. G. DWYER AND C. W. WILKERSON 3.6. Proof. Let W be the Weyl group of X. For p odd, the lemma is a consequence of the fact that the fibration BT ! BN T ! BW has a section [2]; the Serre spectral sequence then gives isomorphisms H Zp2BN T ~= HZp2(B W ) im(H Zp2BT ! H Zp2BNT ) ~= HZp2(B W ) H0(W ; HZp2(B T )) . For p = 2, note that by [14 , 7.4] there are are splittings T ~= T1 x T2, N T ~=N T1 x N T2, W ~=W1 x W2, where N T1 is obtained from the normalizer of the maximal torus in a connected compact Lie group T by F2-completion of the maximal torus [14 , 8.1, 9.1], and NT2 is the product of a number of copies of the normalizer of the torus in the exotic 2-compact group DI(4). As is clear from [14 , 7.2], the group H 0(W2; HZ22BT2) vanishes, and so the image of the map H Z22BT2 ! H Z22BNT2 is trivial. It thus suffices to consider the image of the map H Z22BT1 ! H Z22BNT1, or, equivalently, to show that if G is a connected compact Lie group with maximal torus TG and torus normalizer N TG , the image of the map H 2(B TG ; Z2) ! H 2(B N TG ; Z2) is a summand of H 2(B N TG ; Z2). This follows from 3.3. Recall that a monomorphism Y ! X of p-compact groups [10 , 3.2] is said to be of maximal rank [11 , 4.1] if for some (equivalently, any) maximal torus T of Y the composite T ! Y ! X is a maximal torus for X. 3.7. Lemma. If f : Y ! X is a monomorphism of p-compact groups, then f has maximal rank if and only if O(X=Y ) 6= 0. Proof. Let T be a maximal torus for Y . There is a fibration sequence Y =T ! X=T ! X=Y which by [11 , 10.6] gives a product formula O(X=T ) = O(Y =T )O(X=Y ) (we leave it to the reader to check the nilpotent action condition). The lemma follows from the fact that a maximal torus T of a p-compact group Z is characterized by the fact that O(Z=T ) 6= 0 [11 , 2.15]. The following proposition is the 1-connected case of 1.1. 3.8. Proposition. If X is 1-connected, the natural map H Zp2BT ! H Zp2BNT is zero. Proof. By 3.5 it is enough to show that H 2(B T ; Fp) ! H 2(B N T ; Fp) is zero. Since H 2(B T ; Fp) is detected on maps B Z=p ! B T , it is even enough to show that for any map B Z=p ! BT , the induced composite H 2(B N T ; Fp) ! H 2(B T ; Fp) ! H 2(B Z=p; Fp) is trivial. FUNDAMENTAL GROUP 11 Pick such a map B Z=p ! B T , and consider the following commu- tative diagram of vertical fibration sequences. The space V is defined so that lower square is a fibre square; equivalently, V is the Borel con- struction of the action of T on X= NpT . The map c is the composite BZ=p ! B T ! B X. The space (X= NpT )hZ=p is a homotopy fixed point set [10 , x10] which in this case amounts to the space of sections of the fibration V ! BZ=p. (X= NpT )hZ=p --- ! X= NpT -- - ! X= NpT ? ? ? ? ? ? y y y (X= NpT )hZ=px BZ=p --a-! V -- b-! B NpT ? ? ? u?y v?y w?y B Z=p --=-! BZ=p -- c-! B X If f is a map of spaces, let f* be the induced map on Fp-cohomology. By 2.3, v# b* = c*w# ; since X is 1-connected, v# b* is trivial in dimension 2. Let S H *(B Z=p; Fp) be the multiplicative subset generated by a nonzero class in dimension 2. By Smith theory [10 , 4.11, 5.7], S-1 a* is an isomorphism, and so by 2.4, S-1 v# = S-1 u# a*. Since H*(B Z=p; Fp) ! S-1 H *(B Z=p; Fp) is a monomorphism, it even follows that v# = u# a*. In particular, the map u# a*b* vanishes in dimension 2. Let j be the map BN pT ! BN T , and i the composite BZ=p -c!BT ! BN T . To fin- ish the proof it is enough to show that u# a*b*j* is a nonzero multiple of i*, since it will then follow, as desired, that i* vanishes in dimension 2. Let Y = (X= NpT )hZ=p. By [10 , 4.6] the Euler characteristic of Y is congruent mod p to O(X= NpT ); this latter is nonzero mod p [10 , 9.9 ff] and in fact is equal to the cardinality of W=Wp. There is a fibration Y ! Map (B Z=p, BN pT )! Map (B Z=p, BX)c where Map (B Z=p, BX)c is the component of maps homotopic to c and Map (B Z=p, BN T )consists of maps which cover c up to homotopy. If we interpret c as a homomorphism Z=p ! X then Y can be expressed as a disjoint union a Y = CX (c)=CNpT (fl) , {fl} where {fl} runs over conjugacy classes of homomorphisms Z=p ! N pT which cover c up to conjugacy, and CT (-) denotes the centralizer in T of the image of the indicated map [10 , x5]. For each component {fl} of Y , let ffl: B Z=p ! B NpT be the map obtained by using any point in 12 W. G. DWYER AND C. W. WILKERSON that component to map BZ=p to Y xB Z=p, and then following through with the composite ba. According to 2.5 X (3.9) u# a*b* = O(Yfl)f*fl. {fl}2ss0Y There are two types of such homomorphisms fl. The first type con- sists of those fl with the property that fl(Z=p) is not contained in the torus T in N pT . In this case CNpT (fl) is not a maximal rank subgroup of CX (c), and so by 3.7 the Euler characteristic of the homogeneous space CX (c)=CNpT (fl) is zero. Here we are using the idea of [11 , 3.2] to com- pute CNpT (fl) and combining this with the observation that the action of N pT =T = Wp on (the discrete approximation to) T is faithful [11 , 2.12]. The components of Y corresponding to such maps fl contribute nothing to the sum on the right hand side of 3.9. Consider then maps fl : Z=p ! N pT which lift c and whose im- age lies in T . For any such fl, the composite j . fflis homotopic to i; this is expressed in [12 , 3.4] as the statement that any two homo- morphisms Z=p ! T which are conjugate in X are conjugate in N T . Write ss0Y = A [ B, where A is indexed by lifts of the first kind, and B by liftsPof the second kind. Since O(Yfl) = 0 if {fl} 2 A, it is clear that {fl}2BO(Yfl) = O(Y ) is nonzero mod p. On the other hand, the argument above gives X u# a*b*j* = O(Yfl)i* {fl}2B It follows that u# a*b*j* is a nonzero multiple of i*, as required. Note that any connected covering space of a connected p-compact group is again a p-compact group [21 , 3.3]. 3.10. Lemma. Let X be a connected p-compact group with maximal torus T , and X0 the universal cover of X, with maximal torus T 0. Then the cokernel of ss1T 0! ss1T [10 , 8.11] is isomorphic to the cokernel of ss1X0 ! ss1X. Proof. Let T 00be the identity component of the center of X. According to [21 , 5.4] there is a short exact sequence K ! X0x T 00! X is which K is a finite abelian subgroup of the center of X0. This gives rise to a parallel exact sequence K ! T 0x T 00! T . FUNDAMENTAL GROUP 13 Combining the two leads to the conclusion that the map X0=T 0! X=T is an equivalence. The proof is completed by taking the parallel fibration sequences X0=T 0 --- ! BT 0-- - ! B X0 ? ? ? ~ ?y ?y ?y X=T --- ! BT -- - ! BX and comparing the associated long exact homotopy sequences. Proof of 1.1. Let X0 be the universal cover of X, with maximal torus T 0and torus normalizer N T 0. Consider the commutative diagram HZp2BT 0 --- ! HZp2BN T 0-- - ! H Zp2BX0 (= 0) ? ? ? ? ? ? y y y H Zp2BT --- ! HZp2BN T -- - ! H Zp2BX The homology groups in the outside columns are isomorphic to the corresponding ss2's. By 3.8, the image of the map H Zp2BT 0! H Zp2BNT 0 is zero. To prove the theorem, it is enough to show that if there is an x 2 H Zp2BT which does not map to 0 in H Zp2BN T but does map to 0 in H Zp2BX, then there exists an x0 2 H Zp2BT 0which does not map to 0 in H Zp2BN T 0. But by 3.10 the cokernel of ss2 BT 0! ss2 BT is isomorphic to the cokernel of ss2 BX0 ! ss2 BX, i.e., to ss2 BX, so that any such x 2 H Zp2BT automatically lifts to an x0 2 H Zp2BT 0. By the commutativity of the diagram, x0 has the desired (impossible) property. References [1]J. F. Adams, Lectures on Lie groups, W. A. Benjamin, Inc., New York- Amsterdam, 1969. [2]K. K. S. Andersen, The normalizer splitting conjecture for p-compact groups, Fund. Math. 161 (1999), no. 1-2, 1-16, Algebraic topology (Kazimierz Dolny, 1997). [3]K. K. S. Andersen and J. Grodal, The classification of 2-compact groups, preprint, 2006. [4]K. K. S. Andersen, J. Grodal, J. M. Moller, and A. Viruel, The classificati* *on of p-compact groups for odd p, Annals of Math., to appear. [5]J. C. Becker and D. H. Gottlieb, The transfer map and fiber bundles, Topolo* *gy 14 (1975), 1-12. [6]A. Dold, The fixed point transfer of fibre-preserving maps, Math. Z. 148 (1* *976), no. 3, 215-244. [7]W. G. Dwyer, Transfer maps for fibrations, Math. Proc. Cambridge Philos. Soc. 120 (1996), no. 2, 221-235. 14 W. G. DWYER AND C. W. WILKERSON [8]W. G. Dwyer, J. P. C. Greenlees, and S. Iyengar, Duality in algebra and top* *ol- ogy, Adv. Math. 200 (2006), no. 2, 357-402. [9]W. G. Dwyer and C. W. Wilkerson, Smith theory and the functor T , Comment. Math. Helv. 66 (1991), no. 1, 1-17. [10]______, Homotopy fixed-point methods for Lie groups and finite loop spaces, Ann. of Math. (2) 139 (1994), no. 2, 395-442. [11]______, The center of a p-compact group, The ~Cech centennial (Boston, MA, 1993), Contemp. Math., vol. 181, Amer. Math. Soc., Providence, RI, 1995, pp. 119-157. [12]______, Product splittings for p-compact groups, Fund. Math. 147 (1995), no. 3, 279-300. [13]______, The elementary geometric structure of compact Lie groups, Bull. Lon- don Math. Soc. 30 (1998), no. 4, 337-364. [14]______, Normalizers of tori, Geom. Topol. 9 (2005), 1337-1380 (electronic). [15]A. D. Elmendorf, I. K~r'i~z, M. A. Mandell, and J. P. May, Modern foundatio* *ns for stable homotopy theory, Handbook of algebraic topology, North-Holland, Amsterdam, 1995, pp. 213-253. [16]A. D. Elmendorf, I. Kriz, M. A. Mandell, and J. P. May, Rings, modules, and algebras in stable homotopy theory, Mathematical Surveys and Monographs, vol. 47, American Mathematical Society, Providence, RI, 1997, With an ap- pendix by M. Cole. [17]M. Feshbach, The transfer and compact Lie groups, Trans. Amer. Math. Soc. 251 (1979), 139-169. [18]M. Hovey, B. Shipley, and J. Smith, Symmetric spectra, J. Amer. Math. Soc. 13 (2000), no. 1, 149-208. [19]M. A. Mandell, E1 algebras and p-adic homotopy theory, Topology 40 (2001), no. 1, 43-94. [20]______, Topological Andr'e-Quillen cohomology and E1 Andr'e-Quillen coho- mology, Adv. Math. 177 (2003), no. 2, 227-279. [21]J. M. Moller and D. Notbohm, Centers and finite coverings of finite loop sp* *aces, J. Reine Angew. Math. 456 (1994), 99-133. [22]B. Shipley, HZ-algebra spectra are differential graded algebras, preprint (* *2004). E-mail address: dwyer.1@nd.edu E-mail address: wilker@math.purdue.edu